2009-04-26 03:13:08 +04:00
|
|
|
# dirstate.py - working directory tracking for mercurial
|
|
|
|
#
|
|
|
|
# Copyright 2005-2007 Matt Mackall <mpm@selenic.com>
|
|
|
|
#
|
|
|
|
# This software may be used and distributed according to the terms of the
|
2010-01-20 07:20:08 +03:00
|
|
|
# GNU General Public License version 2 or any later version.
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2008-03-07 00:23:26 +03:00
|
|
|
from node import nullid
|
2006-12-15 05:25:19 +03:00
|
|
|
from i18n import _
|
2013-11-07 03:19:04 +04:00
|
|
|
import scmutil, util, ignore, osutil, parsers, encoding, pathutil
|
2013-02-10 20:23:14 +04:00
|
|
|
import os, stat, errno, gc
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2009-04-30 05:47:18 +04:00
|
|
|
propertycache = util.propertycache
|
2012-03-01 19:42:49 +04:00
|
|
|
filecache = scmutil.filecache
|
2012-10-09 02:50:42 +04:00
|
|
|
_rangemask = 0x7fffffff
|
2012-03-01 19:42:49 +04:00
|
|
|
|
parsers: inline fields of dirstate values in C version
Previously, while unpacking the dirstate we'd create 3-4 new CPython objects
for most dirstate values:
- the state is a single character string, which is pooled by CPython
- the mode is a new object if it isn't 0 due to being in the lookup set
- the size is a new object if it is greater than 255
- the mtime is a new object if it isn't -1 due to being in the lookup set
- the tuple to contain them all
In some cases such as regular hg status, we actually look at all the objects.
In other cases like hg add, hg status for a subdirectory, or hg status with the
third-party hgwatchman enabled, we look at almost none of the objects.
This patch eliminates most object creation in these cases by defining a custom
C struct that is exposed to Python with an interface similar to a tuple. Only
when tuple elements are actually requested are the respective objects created.
The gains, where they're expected, are significant. The following tests are run
against a working copy with over 270,000 files.
parse_dirstate becomes significantly faster:
$ hg perfdirstate
before: wall 0.186437 comb 0.180000 user 0.160000 sys 0.020000 (best of 35)
after: wall 0.093158 comb 0.100000 user 0.090000 sys 0.010000 (best of 95)
and as a result, several commands benefit:
$ time hg status # with hgwatchman enabled
before: 0.42s user 0.14s system 99% cpu 0.563 total
after: 0.34s user 0.12s system 99% cpu 0.471 total
$ time hg add new-file
before: 0.85s user 0.18s system 99% cpu 1.033 total
after: 0.76s user 0.17s system 99% cpu 0.931 total
There is a slight regression in regular status performance, but this is fixed
in an upcoming patch.
2014-05-28 01:27:41 +04:00
|
|
|
dirstatetuple = parsers.dirstatetuple
|
2014-05-28 04:10:28 +04:00
|
|
|
|
2012-03-01 19:42:49 +04:00
|
|
|
class repocache(filecache):
|
|
|
|
"""filecache for files in .hg/"""
|
|
|
|
def join(self, obj, fname):
|
|
|
|
return obj._opener.join(fname)
|
2007-06-18 22:24:34 +04:00
|
|
|
|
2012-03-01 19:49:59 +04:00
|
|
|
class rootcache(filecache):
|
|
|
|
"""filecache for files in the repository root"""
|
|
|
|
def join(self, obj, fname):
|
|
|
|
return obj._join(fname)
|
|
|
|
|
2005-11-19 09:48:47 +03:00
|
|
|
class dirstate(object):
|
2006-06-04 04:25:27 +04:00
|
|
|
|
2010-11-22 21:43:31 +03:00
|
|
|
def __init__(self, opener, ui, root, validate):
|
2009-12-28 01:24:05 +03:00
|
|
|
'''Create a new dirstate object.
|
|
|
|
|
|
|
|
opener is an open()-like callable that can be used to open the
|
|
|
|
dirstate file; root is the root of the directory tracked by
|
|
|
|
the dirstate.
|
|
|
|
'''
|
2007-06-18 22:24:34 +04:00
|
|
|
self._opener = opener
|
2010-11-22 21:43:31 +03:00
|
|
|
self._validate = validate
|
2007-06-18 22:24:34 +04:00
|
|
|
self._root = root
|
2008-09-02 17:12:50 +04:00
|
|
|
self._rootdir = os.path.join(root, '')
|
2007-07-22 01:02:09 +04:00
|
|
|
self._dirty = False
|
2007-07-22 01:44:38 +04:00
|
|
|
self._dirtypl = False
|
2012-01-08 21:15:54 +04:00
|
|
|
self._lastnormaltime = 0
|
2007-06-18 22:24:34 +04:00
|
|
|
self._ui = ui
|
2012-03-01 19:39:58 +04:00
|
|
|
self._filecache = {}
|
2014-09-05 22:34:29 +04:00
|
|
|
self._parentwriters = 0
|
|
|
|
|
|
|
|
def beginparentchange(self):
|
|
|
|
'''Marks the beginning of a set of changes that involve changing
|
|
|
|
the dirstate parents. If there is an exception during this time,
|
|
|
|
the dirstate will not be written when the wlock is released. This
|
|
|
|
prevents writing an incoherent dirstate where the parent doesn't
|
|
|
|
match the contents.
|
|
|
|
'''
|
|
|
|
self._parentwriters += 1
|
|
|
|
|
|
|
|
def endparentchange(self):
|
|
|
|
'''Marks the end of a set of changes that involve changing the
|
|
|
|
dirstate parents. Once all parent changes have been marked done,
|
|
|
|
the wlock will be free to write the dirstate on release.
|
|
|
|
'''
|
|
|
|
if self._parentwriters > 0:
|
|
|
|
self._parentwriters -= 1
|
|
|
|
|
|
|
|
def pendingparentchange(self):
|
|
|
|
'''Returns true if the dirstate is in the middle of a set of changes
|
|
|
|
that modify the dirstate parent.
|
|
|
|
'''
|
|
|
|
return self._parentwriters > 0
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2009-04-30 05:47:18 +04:00
|
|
|
@propertycache
|
|
|
|
def _map(self):
|
2009-10-01 23:36:45 +04:00
|
|
|
'''Return the dirstate contents as a map from filename to
|
|
|
|
(state, mode, size, time).'''
|
2009-04-30 05:47:18 +04:00
|
|
|
self._read()
|
|
|
|
return self._map
|
|
|
|
|
|
|
|
@propertycache
|
|
|
|
def _copymap(self):
|
|
|
|
self._read()
|
|
|
|
return self._copymap
|
|
|
|
|
|
|
|
@propertycache
|
|
|
|
def _foldmap(self):
|
|
|
|
f = {}
|
2013-04-30 00:00:48 +04:00
|
|
|
for name, s in self._map.iteritems():
|
|
|
|
if s[0] != 'r':
|
|
|
|
f[util.normcase(name)] = name
|
2012-03-29 04:24:13 +04:00
|
|
|
for name in self._dirs:
|
|
|
|
f[util.normcase(name)] = name
|
2011-12-16 16:09:40 +04:00
|
|
|
f['.'] = '.' # prevents useless util.fspath() invocation
|
2009-04-30 05:47:18 +04:00
|
|
|
return f
|
|
|
|
|
2012-03-01 19:42:49 +04:00
|
|
|
@repocache('branch')
|
2009-04-30 05:47:18 +04:00
|
|
|
def _branch(self):
|
|
|
|
try:
|
2011-05-02 12:11:18 +04:00
|
|
|
return self._opener.read("branch").strip() or "default"
|
2012-01-06 10:37:59 +04:00
|
|
|
except IOError, inst:
|
|
|
|
if inst.errno != errno.ENOENT:
|
|
|
|
raise
|
2009-04-30 05:47:18 +04:00
|
|
|
return "default"
|
|
|
|
|
|
|
|
@propertycache
|
|
|
|
def _pl(self):
|
|
|
|
try:
|
2010-12-24 17:23:01 +03:00
|
|
|
fp = self._opener("dirstate")
|
|
|
|
st = fp.read(40)
|
|
|
|
fp.close()
|
2009-06-05 01:21:09 +04:00
|
|
|
l = len(st)
|
|
|
|
if l == 40:
|
2009-04-30 05:47:18 +04:00
|
|
|
return st[:20], st[20:40]
|
2009-06-05 01:21:09 +04:00
|
|
|
elif l > 0 and l < 40:
|
2009-05-25 21:48:15 +04:00
|
|
|
raise util.Abort(_('working directory state appears damaged!'))
|
2009-04-30 05:47:18 +04:00
|
|
|
except IOError, err:
|
2010-01-25 09:05:27 +03:00
|
|
|
if err.errno != errno.ENOENT:
|
|
|
|
raise
|
2009-04-30 05:47:18 +04:00
|
|
|
return [nullid, nullid]
|
|
|
|
|
|
|
|
@propertycache
|
|
|
|
def _dirs(self):
|
2013-04-11 02:08:26 +04:00
|
|
|
return scmutil.dirs(self._map, 'r')
|
2009-04-30 05:47:18 +04:00
|
|
|
|
2012-02-22 19:07:54 +04:00
|
|
|
def dirs(self):
|
|
|
|
return self._dirs
|
|
|
|
|
2012-03-01 19:49:59 +04:00
|
|
|
@rootcache('.hgignore')
|
2009-04-30 05:47:18 +04:00
|
|
|
def _ignore(self):
|
|
|
|
files = [self._join('.hgignore')]
|
|
|
|
for name, path in self._ui.configitems("ui"):
|
|
|
|
if name == 'ignore' or name.startswith('ignore.'):
|
2009-10-19 23:19:28 +04:00
|
|
|
files.append(util.expandpath(path))
|
2009-04-30 05:47:18 +04:00
|
|
|
return ignore.ignore(self._root, files, self._ui.warn)
|
|
|
|
|
|
|
|
@propertycache
|
|
|
|
def _slash(self):
|
|
|
|
return self._ui.configbool('ui', 'slash') and os.sep != '/'
|
|
|
|
|
|
|
|
@propertycache
|
|
|
|
def _checklink(self):
|
|
|
|
return util.checklink(self._root)
|
|
|
|
|
|
|
|
@propertycache
|
|
|
|
def _checkexec(self):
|
|
|
|
return util.checkexec(self._root)
|
|
|
|
|
|
|
|
@propertycache
|
|
|
|
def _checkcase(self):
|
|
|
|
return not util.checkcase(self._join('.hg'))
|
|
|
|
|
2007-07-22 01:02:09 +04:00
|
|
|
def _join(self, f):
|
2008-09-02 17:12:50 +04:00
|
|
|
# much faster than os.path.join()
|
2008-09-02 19:32:07 +04:00
|
|
|
# it's safe because f is always a relative path
|
2008-09-02 17:12:50 +04:00
|
|
|
return self._rootdir + f
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2011-10-23 01:12:33 +04:00
|
|
|
def flagfunc(self, buildfallback):
|
|
|
|
if self._checklink and self._checkexec:
|
|
|
|
def f(x):
|
2013-04-03 22:35:27 +04:00
|
|
|
try:
|
|
|
|
st = os.lstat(self._join(x))
|
|
|
|
if util.statislink(st):
|
|
|
|
return 'l'
|
|
|
|
if util.statisexec(st):
|
|
|
|
return 'x'
|
|
|
|
except OSError:
|
|
|
|
pass
|
2011-10-23 01:12:33 +04:00
|
|
|
return ''
|
|
|
|
return f
|
|
|
|
|
|
|
|
fallback = buildfallback()
|
2008-06-26 22:46:34 +04:00
|
|
|
if self._checklink:
|
|
|
|
def f(x):
|
2008-09-02 17:12:50 +04:00
|
|
|
if os.path.islink(self._join(x)):
|
2008-06-26 22:46:34 +04:00
|
|
|
return 'l'
|
|
|
|
if 'x' in fallback(x):
|
|
|
|
return 'x'
|
|
|
|
return ''
|
|
|
|
return f
|
|
|
|
if self._checkexec:
|
|
|
|
def f(x):
|
|
|
|
if 'l' in fallback(x):
|
|
|
|
return 'l'
|
2011-05-08 22:45:47 +04:00
|
|
|
if util.isexec(self._join(x)):
|
2008-06-26 22:46:34 +04:00
|
|
|
return 'x'
|
|
|
|
return ''
|
|
|
|
return f
|
2011-10-23 01:12:33 +04:00
|
|
|
else:
|
|
|
|
return fallback
|
2008-06-26 22:46:34 +04:00
|
|
|
|
2014-02-01 03:13:15 +04:00
|
|
|
@propertycache
|
|
|
|
def _cwd(self):
|
|
|
|
return os.getcwd()
|
|
|
|
|
2005-08-28 01:21:25 +04:00
|
|
|
def getcwd(self):
|
2014-02-01 03:13:15 +04:00
|
|
|
cwd = self._cwd
|
2010-01-25 09:05:27 +03:00
|
|
|
if cwd == self._root:
|
|
|
|
return ''
|
2007-06-18 22:24:34 +04:00
|
|
|
# self._root ends with a path separator if self._root is '/' or 'C:\'
|
|
|
|
rootsep = self._root
|
2008-01-09 15:30:35 +03:00
|
|
|
if not util.endswithsep(rootsep):
|
2007-03-16 06:22:58 +03:00
|
|
|
rootsep += os.sep
|
|
|
|
if cwd.startswith(rootsep):
|
|
|
|
return cwd[len(rootsep):]
|
|
|
|
else:
|
|
|
|
# we're outside the repo. return an absolute path.
|
|
|
|
return cwd
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2007-06-09 06:49:12 +04:00
|
|
|
def pathto(self, f, cwd=None):
|
|
|
|
if cwd is None:
|
|
|
|
cwd = self.getcwd()
|
2007-06-18 22:24:34 +04:00
|
|
|
path = util.pathto(self._root, cwd, f)
|
2007-06-09 06:49:12 +04:00
|
|
|
if self._slash:
|
2013-05-17 23:31:06 +04:00
|
|
|
return util.pconvert(path)
|
2007-06-09 06:49:12 +04:00
|
|
|
return path
|
2007-06-09 06:49:12 +04:00
|
|
|
|
2005-08-28 01:21:25 +04:00
|
|
|
def __getitem__(self, key):
|
2009-10-01 23:36:45 +04:00
|
|
|
'''Return the current state of key (a filename) in the dirstate.
|
2009-12-28 01:24:05 +03:00
|
|
|
|
2009-10-01 23:36:45 +04:00
|
|
|
States are:
|
|
|
|
n normal
|
|
|
|
m needs merging
|
|
|
|
r marked for removal
|
|
|
|
a marked for addition
|
|
|
|
? not tracked
|
|
|
|
'''
|
2007-07-22 01:02:09 +04:00
|
|
|
return self._map.get(key, ("?",))[0]
|
2005-08-28 01:21:25 +04:00
|
|
|
|
|
|
|
def __contains__(self, key):
|
2007-06-18 22:24:34 +04:00
|
|
|
return key in self._map
|
|
|
|
|
|
|
|
def __iter__(self):
|
2009-04-27 01:50:44 +04:00
|
|
|
for x in sorted(self._map):
|
2007-06-18 22:24:34 +04:00
|
|
|
yield x
|
2005-08-28 01:21:25 +04:00
|
|
|
|
completion: add a debugpathcomplete command
The bash_completion code uses "hg status" to generate a list of
possible completions for commands that operate on files in the
working directory. In a large working directory, this can result
in a single tab-completion being very slow (several seconds) as a
result of checking the status of every file, even when there is no
need to check status or no possible matches.
The new debugpathcomplete command gains performance in a few simple
ways:
* Allow completion to operate on just a single directory. When used
to complete the right commands, this considerably reduces the
number of completions returned, at no loss in functionality.
* Never check the status of files. For completions that really must
know if a file is modified, it is faster to use status:
hg status -nm 'glob:myprefix**'
Performance:
Here are the commands used by bash_completion to complete, run in
the root of the mozilla-central working dir (~77,000 files) and
another repo (~165,000 files):
All "normal state" files (used by e.g. remove, revert):
mozilla other
status -nmcd 'glob:**' 1.77 4.10 sec
debugpathcomplete -f -n 0.53 1.26
debugpathcomplete -n 0.17 0.41
("-f" means "complete full paths", rather than the current directory)
Tracked files matching "a":
mozilla other
status -nmcd 'glob:a**' 0.26 0.47
debugpathcomplete -f -n a 0.10 0.24
debugpathcomplete -n a 0.10 0.22
We should be able to further improve completion performance once
the critbit work lands. Right now, our performance is limited by
the need to iterate over all keys in the dirstate.
2013-03-22 03:31:28 +04:00
|
|
|
def iteritems(self):
|
|
|
|
return self._map.iteritems()
|
|
|
|
|
2005-08-28 01:21:25 +04:00
|
|
|
def parents(self):
|
2010-11-22 21:43:31 +03:00
|
|
|
return [self._validate(p) for p in self._pl]
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2011-04-05 00:52:55 +04:00
|
|
|
def p1(self):
|
|
|
|
return self._validate(self._pl[0])
|
|
|
|
|
|
|
|
def p2(self):
|
|
|
|
return self._validate(self._pl[1])
|
|
|
|
|
2007-03-14 02:50:02 +03:00
|
|
|
def branch(self):
|
2010-11-25 00:56:32 +03:00
|
|
|
return encoding.tolocal(self._branch)
|
2007-03-14 02:50:02 +03:00
|
|
|
|
2005-08-28 01:21:25 +04:00
|
|
|
def setparents(self, p1, p2=nullid):
|
2012-04-30 00:25:55 +04:00
|
|
|
"""Set dirstate parents to p1 and p2.
|
|
|
|
|
|
|
|
When moving from two parents to one, 'm' merged entries a
|
|
|
|
adjusted to normal and previous copy records discarded and
|
|
|
|
returned by the call.
|
|
|
|
|
|
|
|
See localrepo.setparents()
|
|
|
|
"""
|
2014-09-05 22:37:44 +04:00
|
|
|
if self._parentwriters == 0:
|
|
|
|
raise ValueError("cannot set dirstate parent without " +
|
|
|
|
" calling dirstate.begingparentchange")
|
|
|
|
|
2007-07-22 01:44:38 +04:00
|
|
|
self._dirty = self._dirtypl = True
|
rebase: skip resolved but emptied revisions
When rebasing, if a conflict occurs and is resolved in a way the rebased
revision becomes empty, it is not skipped, unlike revisions being emptied
without conflicts.
The reason is:
- File 'x' is merged and resolved, merge.update() marks it as 'm' in the
dirstate.
- rebase.concludenode() calls localrepo.commit(), which calls
localrepo.status() which calls dirstate.status(). 'x' shows up as 'm' and is
unconditionnally added to the modified files list, instead of being checked
again.
- localrepo.commit() detects 'x' as changed an create a new revision where only
the manifest parents and linkrev differ.
Marking 'x' as modified without checking it makes sense for regular merges. But
in rebase case, the merge looks normal but the second parent is usually
discarded. When this happens, 'm' files in dirstate are a bit irrelevant and
should be considered 'n' possibly dirty instead. That is what the current patch
does.
Another approach, maybe more efficient, would be to pass another flag to
merge.update() saying the 'branchmerge' is a bit of a lie and recordupdate()
should call dirstate.normallookup() instead of merge().
It is also tempting to add this logic to dirstate.setparents(), moving from two
to one parent is what invalidates the 'm' markers. But this is a far bigger
change to make.
v2: succumb to the temptation and move the logic in dirstate.setparents(). mpm
suggested trying _filecommit() first but it is called by commitctx() which
knows nothing about the dirstate and comes too late into the game. A second
approach was to rewrite the 'm' state into 'n' on the fly in dirstate.status()
which failed for graft in the following case:
$ hg init repo
$ cd repo
$ echo a > a
$ hg ci -qAm0
$ echo a >> a
$ hg ci -m1
$ hg up 0
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ hg mv a b
$ echo c > b
$ hg ci -m2
created new head
$ hg graft 1 --tool internal:local
grafting revision 1
$ hg --config extensions.graphlog= glog --template '{rev} {desc|firstline}\n'
@ 3 1
|
o 2 2
|
| o 1 1
|/
o 0 0
$ hg log -r 3 --debug --patch --git --copies
changeset: 3:19cd7d1417952af13161b94c32e901769104560c
tag: tip
phase: draft
parent: 2:b5c505595c9e9a12d5dd457919c143e05fc16fb8
parent: -1:0000000000000000000000000000000000000000
manifest: 3:3d27ce8d02241aa59b60804805edf103c5c0cda4
user: test
date: Thu Jan 01 00:00:00 1970 +0000
extra: branch=default
extra: source=a03df74c41413a75c0a42997fc36c2de97b26658
description:
1
Here, revision 3 is created because there is a copy record for 'b' in the
dirstate and thus 'b' is considered modified. But this information is discarded
at commit time since 'b' content is unchanged. I do not know if discarding this
information is correct or not, but at this time we cannot represent it anyway.
This patch therefore implements the last solution of moving the logic into
dirstate.setparents(). It does not sound crazy as 'm' files makes no sense with
only one parent. It also makes dirstate.merge() calls .lookupnormal() if there
is one parent, to preserve the invariant.
I am a bit concerned about introducing this kind of stateful behaviour to
existing code which historically treated setparents() as a basic setter without
side-effects. And doing that during the code freeze.
2012-04-22 22:06:36 +04:00
|
|
|
oldp2 = self._pl[1]
|
2007-06-18 22:24:34 +04:00
|
|
|
self._pl = p1, p2
|
2012-04-30 00:25:55 +04:00
|
|
|
copies = {}
|
rebase: skip resolved but emptied revisions
When rebasing, if a conflict occurs and is resolved in a way the rebased
revision becomes empty, it is not skipped, unlike revisions being emptied
without conflicts.
The reason is:
- File 'x' is merged and resolved, merge.update() marks it as 'm' in the
dirstate.
- rebase.concludenode() calls localrepo.commit(), which calls
localrepo.status() which calls dirstate.status(). 'x' shows up as 'm' and is
unconditionnally added to the modified files list, instead of being checked
again.
- localrepo.commit() detects 'x' as changed an create a new revision where only
the manifest parents and linkrev differ.
Marking 'x' as modified without checking it makes sense for regular merges. But
in rebase case, the merge looks normal but the second parent is usually
discarded. When this happens, 'm' files in dirstate are a bit irrelevant and
should be considered 'n' possibly dirty instead. That is what the current patch
does.
Another approach, maybe more efficient, would be to pass another flag to
merge.update() saying the 'branchmerge' is a bit of a lie and recordupdate()
should call dirstate.normallookup() instead of merge().
It is also tempting to add this logic to dirstate.setparents(), moving from two
to one parent is what invalidates the 'm' markers. But this is a far bigger
change to make.
v2: succumb to the temptation and move the logic in dirstate.setparents(). mpm
suggested trying _filecommit() first but it is called by commitctx() which
knows nothing about the dirstate and comes too late into the game. A second
approach was to rewrite the 'm' state into 'n' on the fly in dirstate.status()
which failed for graft in the following case:
$ hg init repo
$ cd repo
$ echo a > a
$ hg ci -qAm0
$ echo a >> a
$ hg ci -m1
$ hg up 0
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ hg mv a b
$ echo c > b
$ hg ci -m2
created new head
$ hg graft 1 --tool internal:local
grafting revision 1
$ hg --config extensions.graphlog= glog --template '{rev} {desc|firstline}\n'
@ 3 1
|
o 2 2
|
| o 1 1
|/
o 0 0
$ hg log -r 3 --debug --patch --git --copies
changeset: 3:19cd7d1417952af13161b94c32e901769104560c
tag: tip
phase: draft
parent: 2:b5c505595c9e9a12d5dd457919c143e05fc16fb8
parent: -1:0000000000000000000000000000000000000000
manifest: 3:3d27ce8d02241aa59b60804805edf103c5c0cda4
user: test
date: Thu Jan 01 00:00:00 1970 +0000
extra: branch=default
extra: source=a03df74c41413a75c0a42997fc36c2de97b26658
description:
1
Here, revision 3 is created because there is a copy record for 'b' in the
dirstate and thus 'b' is considered modified. But this information is discarded
at commit time since 'b' content is unchanged. I do not know if discarding this
information is correct or not, but at this time we cannot represent it anyway.
This patch therefore implements the last solution of moving the logic into
dirstate.setparents(). It does not sound crazy as 'm' files makes no sense with
only one parent. It also makes dirstate.merge() calls .lookupnormal() if there
is one parent, to preserve the invariant.
I am a bit concerned about introducing this kind of stateful behaviour to
existing code which historically treated setparents() as a basic setter without
side-effects. And doing that during the code freeze.
2012-04-22 22:06:36 +04:00
|
|
|
if oldp2 != nullid and p2 == nullid:
|
|
|
|
# Discard 'm' markers when moving away from a merge state
|
|
|
|
for f, s in self._map.iteritems():
|
|
|
|
if s[0] == 'm':
|
2012-04-30 00:25:55 +04:00
|
|
|
if f in self._copymap:
|
|
|
|
copies[f] = self._copymap[f]
|
rebase: skip resolved but emptied revisions
When rebasing, if a conflict occurs and is resolved in a way the rebased
revision becomes empty, it is not skipped, unlike revisions being emptied
without conflicts.
The reason is:
- File 'x' is merged and resolved, merge.update() marks it as 'm' in the
dirstate.
- rebase.concludenode() calls localrepo.commit(), which calls
localrepo.status() which calls dirstate.status(). 'x' shows up as 'm' and is
unconditionnally added to the modified files list, instead of being checked
again.
- localrepo.commit() detects 'x' as changed an create a new revision where only
the manifest parents and linkrev differ.
Marking 'x' as modified without checking it makes sense for regular merges. But
in rebase case, the merge looks normal but the second parent is usually
discarded. When this happens, 'm' files in dirstate are a bit irrelevant and
should be considered 'n' possibly dirty instead. That is what the current patch
does.
Another approach, maybe more efficient, would be to pass another flag to
merge.update() saying the 'branchmerge' is a bit of a lie and recordupdate()
should call dirstate.normallookup() instead of merge().
It is also tempting to add this logic to dirstate.setparents(), moving from two
to one parent is what invalidates the 'm' markers. But this is a far bigger
change to make.
v2: succumb to the temptation and move the logic in dirstate.setparents(). mpm
suggested trying _filecommit() first but it is called by commitctx() which
knows nothing about the dirstate and comes too late into the game. A second
approach was to rewrite the 'm' state into 'n' on the fly in dirstate.status()
which failed for graft in the following case:
$ hg init repo
$ cd repo
$ echo a > a
$ hg ci -qAm0
$ echo a >> a
$ hg ci -m1
$ hg up 0
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ hg mv a b
$ echo c > b
$ hg ci -m2
created new head
$ hg graft 1 --tool internal:local
grafting revision 1
$ hg --config extensions.graphlog= glog --template '{rev} {desc|firstline}\n'
@ 3 1
|
o 2 2
|
| o 1 1
|/
o 0 0
$ hg log -r 3 --debug --patch --git --copies
changeset: 3:19cd7d1417952af13161b94c32e901769104560c
tag: tip
phase: draft
parent: 2:b5c505595c9e9a12d5dd457919c143e05fc16fb8
parent: -1:0000000000000000000000000000000000000000
manifest: 3:3d27ce8d02241aa59b60804805edf103c5c0cda4
user: test
date: Thu Jan 01 00:00:00 1970 +0000
extra: branch=default
extra: source=a03df74c41413a75c0a42997fc36c2de97b26658
description:
1
Here, revision 3 is created because there is a copy record for 'b' in the
dirstate and thus 'b' is considered modified. But this information is discarded
at commit time since 'b' content is unchanged. I do not know if discarding this
information is correct or not, but at this time we cannot represent it anyway.
This patch therefore implements the last solution of moving the logic into
dirstate.setparents(). It does not sound crazy as 'm' files makes no sense with
only one parent. It also makes dirstate.merge() calls .lookupnormal() if there
is one parent, to preserve the invariant.
I am a bit concerned about introducing this kind of stateful behaviour to
existing code which historically treated setparents() as a basic setter without
side-effects. And doing that during the code freeze.
2012-04-22 22:06:36 +04:00
|
|
|
self.normallookup(f)
|
2012-04-30 00:25:55 +04:00
|
|
|
return copies
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2007-03-14 02:50:02 +03:00
|
|
|
def setbranch(self, branch):
|
2010-11-25 00:56:32 +03:00
|
|
|
self._branch = encoding.fromlocal(branch)
|
2012-04-19 19:11:42 +04:00
|
|
|
f = self._opener('branch', 'w', atomictemp=True)
|
|
|
|
try:
|
|
|
|
f.write(self._branch + '\n')
|
|
|
|
f.close()
|
2012-12-16 22:33:00 +04:00
|
|
|
|
|
|
|
# make sure filecache has the correct stat info for _branch after
|
|
|
|
# replacing the underlying file
|
|
|
|
ce = self._filecache['_branch']
|
|
|
|
if ce:
|
|
|
|
ce.refresh()
|
2012-12-15 22:19:07 +04:00
|
|
|
except: # re-raises
|
|
|
|
f.discard()
|
|
|
|
raise
|
2007-03-14 02:50:02 +03:00
|
|
|
|
2007-06-18 22:24:34 +04:00
|
|
|
def _read(self):
|
2007-06-18 22:24:34 +04:00
|
|
|
self._map = {}
|
|
|
|
self._copymap = {}
|
2007-06-18 22:24:33 +04:00
|
|
|
try:
|
2011-05-02 12:11:18 +04:00
|
|
|
st = self._opener.read("dirstate")
|
2007-06-18 22:24:33 +04:00
|
|
|
except IOError, err:
|
2010-01-25 09:05:27 +03:00
|
|
|
if err.errno != errno.ENOENT:
|
|
|
|
raise
|
2007-06-18 22:24:33 +04:00
|
|
|
return
|
|
|
|
if not st:
|
|
|
|
return
|
|
|
|
|
2013-02-10 20:23:14 +04:00
|
|
|
# Python's garbage collector triggers a GC each time a certain number
|
|
|
|
# of container objects (the number being defined by
|
|
|
|
# gc.get_threshold()) are allocated. parse_dirstate creates a tuple
|
|
|
|
# for each file in the dirstate. The C version then immediately marks
|
|
|
|
# them as not to be tracked by the collector. However, this has no
|
|
|
|
# effect on when GCs are triggered, only on what objects the GC looks
|
|
|
|
# into. This means that O(number of files) GCs are unavoidable.
|
|
|
|
# Depending on when in the process's lifetime the dirstate is parsed,
|
|
|
|
# this can get very expensive. As a workaround, disable GC while
|
|
|
|
# parsing the dirstate.
|
|
|
|
gcenabled = gc.isenabled()
|
|
|
|
gc.disable()
|
|
|
|
try:
|
|
|
|
p = parsers.parse_dirstate(self._map, self._copymap, st)
|
|
|
|
finally:
|
|
|
|
if gcenabled:
|
|
|
|
gc.enable()
|
2007-07-20 02:43:25 +04:00
|
|
|
if not self._dirtypl:
|
2008-10-13 00:21:08 +04:00
|
|
|
self._pl = p
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2007-06-18 22:24:34 +04:00
|
|
|
def invalidate(self):
|
2010-12-02 05:43:06 +03:00
|
|
|
for a in ("_map", "_copymap", "_foldmap", "_branch", "_pl", "_dirs",
|
|
|
|
"_ignore"):
|
2007-07-20 02:43:25 +04:00
|
|
|
if a in self.__dict__:
|
|
|
|
delattr(self, a)
|
2012-01-08 21:15:54 +04:00
|
|
|
self._lastnormaltime = 0
|
2007-07-22 01:02:09 +04:00
|
|
|
self._dirty = False
|
2014-09-05 22:34:29 +04:00
|
|
|
self._parentwriters = 0
|
2007-04-24 23:02:51 +04:00
|
|
|
|
2005-08-28 01:21:25 +04:00
|
|
|
def copy(self, source, dest):
|
2009-12-28 01:24:05 +03:00
|
|
|
"""Mark dest as a copy of source. Unmark dest if source is None."""
|
2008-06-15 15:01:03 +04:00
|
|
|
if source == dest:
|
|
|
|
return
|
2007-07-22 01:02:09 +04:00
|
|
|
self._dirty = True
|
2009-01-04 23:32:40 +03:00
|
|
|
if source is not None:
|
|
|
|
self._copymap[dest] = source
|
|
|
|
elif dest in self._copymap:
|
|
|
|
del self._copymap[dest]
|
2005-08-28 01:21:25 +04:00
|
|
|
|
|
|
|
def copied(self, file):
|
2007-06-18 22:24:34 +04:00
|
|
|
return self._copymap.get(file, None)
|
2006-09-26 02:53:17 +04:00
|
|
|
|
|
|
|
def copies(self):
|
2007-06-18 22:24:34 +04:00
|
|
|
return self._copymap
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2008-07-12 03:46:02 +04:00
|
|
|
def _droppath(self, f):
|
|
|
|
if self[f] not in "?r" and "_dirs" in self.__dict__:
|
2013-04-11 02:08:26 +04:00
|
|
|
self._dirs.delpath(f)
|
2007-06-18 22:24:34 +04:00
|
|
|
|
2012-07-04 03:31:37 +04:00
|
|
|
def _addpath(self, f, state, mode, size, mtime):
|
2007-11-05 20:05:44 +03:00
|
|
|
oldstate = self[f]
|
2012-07-04 03:31:37 +04:00
|
|
|
if state == 'a' or oldstate == 'r':
|
2011-04-21 15:18:52 +04:00
|
|
|
scmutil.checkfilename(f)
|
2008-07-12 03:46:02 +04:00
|
|
|
if f in self._dirs:
|
|
|
|
raise util.Abort(_('directory %r already in dirstate') % f)
|
|
|
|
# shadows
|
2013-04-11 02:08:25 +04:00
|
|
|
for d in scmutil.finddirs(f):
|
2008-07-12 03:46:02 +04:00
|
|
|
if d in self._dirs:
|
|
|
|
break
|
|
|
|
if d in self._map and self[d] != 'r':
|
|
|
|
raise util.Abort(
|
|
|
|
_('file %r in dirstate clashes with %r') % (d, f))
|
|
|
|
if oldstate in "?r" and "_dirs" in self.__dict__:
|
2013-04-11 02:08:26 +04:00
|
|
|
self._dirs.addpath(f)
|
2012-06-18 19:06:42 +04:00
|
|
|
self._dirty = True
|
2014-05-28 04:10:28 +04:00
|
|
|
self._map[f] = dirstatetuple(state, mode, size, mtime)
|
2007-11-05 20:05:44 +03:00
|
|
|
|
2007-07-22 01:02:09 +04:00
|
|
|
def normal(self, f):
|
2009-12-28 01:24:05 +03:00
|
|
|
'''Mark a file normal and clean.'''
|
2007-07-22 01:02:09 +04:00
|
|
|
s = os.lstat(self._join(f))
|
2011-03-24 20:39:54 +03:00
|
|
|
mtime = int(s.st_mtime)
|
2012-10-09 02:50:42 +04:00
|
|
|
self._addpath(f, 'n', s.st_mode,
|
|
|
|
s.st_size & _rangemask, mtime & _rangemask)
|
2008-01-20 16:39:25 +03:00
|
|
|
if f in self._copymap:
|
2007-07-22 01:02:09 +04:00
|
|
|
del self._copymap[f]
|
2011-03-25 17:03:53 +03:00
|
|
|
if mtime > self._lastnormaltime:
|
|
|
|
# Remember the most recent modification timeslot for status(),
|
2011-03-24 20:39:54 +03:00
|
|
|
# to make sure we won't miss future size-preserving file content
|
|
|
|
# modifications that happen within the same timeslot.
|
2011-03-25 17:03:53 +03:00
|
|
|
self._lastnormaltime = mtime
|
dirstate: avoid a race with multiple commits in the same process
(issue2264, issue2516)
The race happens when two commits in a row change the same file
without changing its size, *if* those two commits happen in the same
second in the same process while holding the same repo lock. For
example:
commit 1:
M a
M b
commit 2: # same process, same second, same repo lock
M b # modify b without changing its size
M c
This first manifested in transplant, which is the most common way to
do multiple commits in the same process. But it can manifest in any
script or extension that does multiple commits under the same repo
lock. (Thus, the test script tests both transplant and a custom script.)
The problem was that dirstate.status() failed to notice the change to
b when localrepo is about to do the second commit, meaning that change
gets left in the working directory. In the context of transplant, that
means either a crash ("RuntimeError: nothing committed after
transplant") or a silently inaccurate transplant, depending on whether
any other files were modified by the second transplanted changeset.
The fix is to make status() work a little harder when we have
previously marked files as clean (state 'normal') in the same process.
Specifically, dirstate.normal() adds files to self._lastnormal, and
other state-changing methods remove them. Then dirstate.status() puts
any files in self._lastnormal into state 'lookup', which will make
localrepository.status() read file contents to see if it has really
changed. So we pay a small performance penalty for the second (and
subsequent) commits in the same process, without affecting the common
case. Anything that does lots of status updates and checks in the
same process could suffer a performance hit.
Incidentally, there is a simpler fix: call dirstate.normallookup() on
every file updated by commit() at the end of the commit. The trouble
with that solution is that it imposes a performance penalty on the
common case: it means the next status-dependent hg command after every
"hg commit" will be a little bit slower. The patch here is more
complex, but only affects performance for the uncommon case.
2011-03-21 00:41:09 +03:00
|
|
|
|
merge: forcefully mark files that we get from the second parent as dirty
After a hg merge, we want to include in the commit all the files that we
got from the second parent, so that we have the correct file-level
history. To make them visible to hg commit, we try to mark them as dirty.
Unfortunately, right now we can't really mark them as dirty[1] - the
best we can do is to mark them as needing a full comparison of their
contents, but they will still be considered clean if they happen to be
identical to the version in the first parent.
This changeset extends the dirstate format in a compatible way, so that
we can mark a file as dirty:
Right now we use a negative file size to indicate we don't have valid
stat data for this entry. In practice, this size is always -1.
This patch uses -2 to indicate that the entry is dirty. Older versions
of hg won't choke on this dirstate, but they may happily mark the file
as clean after a full comparison, destroying all of our hard work.
The patch adds a dirstate.normallookup method with the semantics of the
current normaldirty, and changes normaldirty to forcefully mark the
entry as dirty.
This should fix issue522.
[1] - well, we could put them in state 'm', but that state has a
different meaning.
2007-08-23 08:48:29 +04:00
|
|
|
def normallookup(self, f):
|
2009-12-28 01:24:05 +03:00
|
|
|
'''Mark a file normal, but possibly dirty.'''
|
2008-03-18 10:07:39 +03:00
|
|
|
if self._pl[1] != nullid and f in self._map:
|
|
|
|
# if there is a merge going on and the file was either
|
2010-04-20 13:17:01 +04:00
|
|
|
# in state 'm' (-1) or coming from other parent (-2) before
|
|
|
|
# being removed, restore that state.
|
2008-03-18 10:07:39 +03:00
|
|
|
entry = self._map[f]
|
|
|
|
if entry[0] == 'r' and entry[2] in (-1, -2):
|
|
|
|
source = self._copymap.get(f)
|
|
|
|
if entry[2] == -1:
|
|
|
|
self.merge(f)
|
|
|
|
elif entry[2] == -2:
|
2010-04-20 13:17:01 +04:00
|
|
|
self.otherparent(f)
|
2008-03-18 10:07:39 +03:00
|
|
|
if source:
|
|
|
|
self.copy(source, f)
|
|
|
|
return
|
|
|
|
if entry[0] == 'm' or entry[0] == 'n' and entry[2] == -2:
|
|
|
|
return
|
2012-06-18 19:06:42 +04:00
|
|
|
self._addpath(f, 'n', 0, -1, -1)
|
merge: forcefully mark files that we get from the second parent as dirty
After a hg merge, we want to include in the commit all the files that we
got from the second parent, so that we have the correct file-level
history. To make them visible to hg commit, we try to mark them as dirty.
Unfortunately, right now we can't really mark them as dirty[1] - the
best we can do is to mark them as needing a full comparison of their
contents, but they will still be considered clean if they happen to be
identical to the version in the first parent.
This changeset extends the dirstate format in a compatible way, so that
we can mark a file as dirty:
Right now we use a negative file size to indicate we don't have valid
stat data for this entry. In practice, this size is always -1.
This patch uses -2 to indicate that the entry is dirty. Older versions
of hg won't choke on this dirstate, but they may happily mark the file
as clean after a full comparison, destroying all of our hard work.
The patch adds a dirstate.normallookup method with the semantics of the
current normaldirty, and changes normaldirty to forcefully mark the
entry as dirty.
This should fix issue522.
[1] - well, we could put them in state 'm', but that state has a
different meaning.
2007-08-23 08:48:29 +04:00
|
|
|
if f in self._copymap:
|
|
|
|
del self._copymap[f]
|
|
|
|
|
2010-04-20 13:17:01 +04:00
|
|
|
def otherparent(self, f):
|
|
|
|
'''Mark as coming from the other parent, always dirty.'''
|
|
|
|
if self._pl[1] == nullid:
|
|
|
|
raise util.Abort(_("setting %r to other parent "
|
|
|
|
"only allowed in merges") % f)
|
2012-06-18 19:06:42 +04:00
|
|
|
self._addpath(f, 'n', 0, -2, -1)
|
2007-07-22 01:02:09 +04:00
|
|
|
if f in self._copymap:
|
|
|
|
del self._copymap[f]
|
|
|
|
|
|
|
|
def add(self, f):
|
2009-12-28 01:24:05 +03:00
|
|
|
'''Mark a file added.'''
|
2012-07-04 03:31:37 +04:00
|
|
|
self._addpath(f, 'a', 0, -1, -1)
|
2007-07-22 01:02:09 +04:00
|
|
|
if f in self._copymap:
|
|
|
|
del self._copymap[f]
|
|
|
|
|
|
|
|
def remove(self, f):
|
2009-12-28 01:24:05 +03:00
|
|
|
'''Mark a file removed.'''
|
2007-07-22 01:02:09 +04:00
|
|
|
self._dirty = True
|
2008-07-12 03:46:02 +04:00
|
|
|
self._droppath(f)
|
2008-03-18 10:07:39 +03:00
|
|
|
size = 0
|
|
|
|
if self._pl[1] != nullid and f in self._map:
|
2010-04-20 13:17:01 +04:00
|
|
|
# backup the previous state
|
2008-03-18 10:07:39 +03:00
|
|
|
entry = self._map[f]
|
2010-04-20 13:17:01 +04:00
|
|
|
if entry[0] == 'm': # merge
|
2008-03-18 10:07:39 +03:00
|
|
|
size = -1
|
2010-04-20 13:17:01 +04:00
|
|
|
elif entry[0] == 'n' and entry[2] == -2: # other parent
|
2008-03-18 10:07:39 +03:00
|
|
|
size = -2
|
2014-05-28 04:10:28 +04:00
|
|
|
self._map[f] = dirstatetuple('r', 0, size, 0)
|
2008-03-18 10:07:39 +03:00
|
|
|
if size == 0 and f in self._copymap:
|
2007-07-22 01:02:09 +04:00
|
|
|
del self._copymap[f]
|
|
|
|
|
|
|
|
def merge(self, f):
|
2009-12-28 01:24:05 +03:00
|
|
|
'''Mark a file merged.'''
|
rebase: skip resolved but emptied revisions
When rebasing, if a conflict occurs and is resolved in a way the rebased
revision becomes empty, it is not skipped, unlike revisions being emptied
without conflicts.
The reason is:
- File 'x' is merged and resolved, merge.update() marks it as 'm' in the
dirstate.
- rebase.concludenode() calls localrepo.commit(), which calls
localrepo.status() which calls dirstate.status(). 'x' shows up as 'm' and is
unconditionnally added to the modified files list, instead of being checked
again.
- localrepo.commit() detects 'x' as changed an create a new revision where only
the manifest parents and linkrev differ.
Marking 'x' as modified without checking it makes sense for regular merges. But
in rebase case, the merge looks normal but the second parent is usually
discarded. When this happens, 'm' files in dirstate are a bit irrelevant and
should be considered 'n' possibly dirty instead. That is what the current patch
does.
Another approach, maybe more efficient, would be to pass another flag to
merge.update() saying the 'branchmerge' is a bit of a lie and recordupdate()
should call dirstate.normallookup() instead of merge().
It is also tempting to add this logic to dirstate.setparents(), moving from two
to one parent is what invalidates the 'm' markers. But this is a far bigger
change to make.
v2: succumb to the temptation and move the logic in dirstate.setparents(). mpm
suggested trying _filecommit() first but it is called by commitctx() which
knows nothing about the dirstate and comes too late into the game. A second
approach was to rewrite the 'm' state into 'n' on the fly in dirstate.status()
which failed for graft in the following case:
$ hg init repo
$ cd repo
$ echo a > a
$ hg ci -qAm0
$ echo a >> a
$ hg ci -m1
$ hg up 0
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ hg mv a b
$ echo c > b
$ hg ci -m2
created new head
$ hg graft 1 --tool internal:local
grafting revision 1
$ hg --config extensions.graphlog= glog --template '{rev} {desc|firstline}\n'
@ 3 1
|
o 2 2
|
| o 1 1
|/
o 0 0
$ hg log -r 3 --debug --patch --git --copies
changeset: 3:19cd7d1417952af13161b94c32e901769104560c
tag: tip
phase: draft
parent: 2:b5c505595c9e9a12d5dd457919c143e05fc16fb8
parent: -1:0000000000000000000000000000000000000000
manifest: 3:3d27ce8d02241aa59b60804805edf103c5c0cda4
user: test
date: Thu Jan 01 00:00:00 1970 +0000
extra: branch=default
extra: source=a03df74c41413a75c0a42997fc36c2de97b26658
description:
1
Here, revision 3 is created because there is a copy record for 'b' in the
dirstate and thus 'b' is considered modified. But this information is discarded
at commit time since 'b' content is unchanged. I do not know if discarding this
information is correct or not, but at this time we cannot represent it anyway.
This patch therefore implements the last solution of moving the logic into
dirstate.setparents(). It does not sound crazy as 'm' files makes no sense with
only one parent. It also makes dirstate.merge() calls .lookupnormal() if there
is one parent, to preserve the invariant.
I am a bit concerned about introducing this kind of stateful behaviour to
existing code which historically treated setparents() as a basic setter without
side-effects. And doing that during the code freeze.
2012-04-22 22:06:36 +04:00
|
|
|
if self._pl[1] == nullid:
|
|
|
|
return self.normallookup(f)
|
2007-07-22 01:02:09 +04:00
|
|
|
s = os.lstat(self._join(f))
|
2012-10-09 02:50:42 +04:00
|
|
|
self._addpath(f, 'm', s.st_mode,
|
|
|
|
s.st_size & _rangemask, int(s.st_mtime) & _rangemask)
|
2007-07-22 01:02:09 +04:00
|
|
|
if f in self._copymap:
|
|
|
|
del self._copymap[f]
|
|
|
|
|
2011-05-27 02:15:35 +04:00
|
|
|
def drop(self, f):
|
|
|
|
'''Drop a file from the dirstate'''
|
2011-11-02 00:19:37 +04:00
|
|
|
if f in self._map:
|
|
|
|
self._dirty = True
|
|
|
|
self._droppath(f)
|
|
|
|
del self._map[f]
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2012-04-28 22:29:21 +04:00
|
|
|
def _normalize(self, path, isknown, ignoremissing=False, exists=None):
|
2011-11-16 00:25:11 +04:00
|
|
|
normed = util.normcase(path)
|
2011-03-22 19:59:43 +03:00
|
|
|
folded = self._foldmap.get(normed, None)
|
|
|
|
if folded is None:
|
2012-04-28 22:29:21 +04:00
|
|
|
if isknown:
|
2011-03-22 19:59:43 +03:00
|
|
|
folded = path
|
2008-10-01 01:23:08 +04:00
|
|
|
else:
|
2012-04-28 22:29:21 +04:00
|
|
|
if exists is None:
|
|
|
|
exists = os.path.lexists(os.path.join(self._root, path))
|
|
|
|
if not exists:
|
|
|
|
# Maybe a path component exists
|
|
|
|
if not ignoremissing and '/' in path:
|
|
|
|
d, f = path.rsplit('/', 1)
|
|
|
|
d = self._normalize(d, isknown, ignoremissing, None)
|
|
|
|
folded = d + "/" + f
|
|
|
|
else:
|
|
|
|
# No path components, preserve original case
|
|
|
|
folded = path
|
2012-03-29 04:24:13 +04:00
|
|
|
else:
|
2012-04-28 22:29:21 +04:00
|
|
|
# recursively normalize leading directory components
|
|
|
|
# against dirstate
|
|
|
|
if '/' in normed:
|
|
|
|
d, f = normed.rsplit('/', 1)
|
|
|
|
d = self._normalize(d, isknown, ignoremissing, True)
|
|
|
|
r = self._root + "/" + d
|
|
|
|
folded = d + "/" + util.fspath(f, r)
|
|
|
|
else:
|
|
|
|
folded = util.fspath(normed, self._root)
|
|
|
|
self._foldmap[normed] = folded
|
2012-03-29 04:24:13 +04:00
|
|
|
|
2011-03-22 19:59:43 +03:00
|
|
|
return folded
|
|
|
|
|
2012-04-28 22:29:21 +04:00
|
|
|
def normalize(self, path, isknown=False, ignoremissing=False):
|
2011-03-22 19:59:43 +03:00
|
|
|
'''
|
|
|
|
normalize the case of a pathname when on a casefolding filesystem
|
|
|
|
|
|
|
|
isknown specifies whether the filename came from walking the
|
2012-04-28 22:29:21 +04:00
|
|
|
disk, to avoid extra filesystem access.
|
|
|
|
|
|
|
|
If ignoremissing is True, missing path are returned
|
|
|
|
unchanged. Otherwise, we try harder to normalize possibly
|
|
|
|
existing path components.
|
2011-03-22 19:59:43 +03:00
|
|
|
|
|
|
|
The normalized case is determined based on the following precedence:
|
|
|
|
|
|
|
|
- version of name already stored in the dirstate
|
|
|
|
- version of name stored on disk
|
|
|
|
- version provided via command arguments
|
|
|
|
'''
|
|
|
|
|
|
|
|
if self._checkcase:
|
2012-04-28 22:29:21 +04:00
|
|
|
return self._normalize(path, isknown, ignoremissing)
|
2011-03-22 19:59:43 +03:00
|
|
|
return path
|
2008-06-06 22:23:29 +04:00
|
|
|
|
2007-08-06 06:04:56 +04:00
|
|
|
def clear(self):
|
|
|
|
self._map = {}
|
2007-11-05 20:05:44 +03:00
|
|
|
if "_dirs" in self.__dict__:
|
2010-02-08 17:36:34 +03:00
|
|
|
delattr(self, "_dirs")
|
2007-08-06 06:04:56 +04:00
|
|
|
self._copymap = {}
|
|
|
|
self._pl = [nullid, nullid]
|
2012-01-08 21:15:54 +04:00
|
|
|
self._lastnormaltime = 0
|
2007-08-06 08:00:10 +04:00
|
|
|
self._dirty = True
|
2007-08-06 06:04:56 +04:00
|
|
|
|
2013-03-07 08:13:09 +04:00
|
|
|
def rebuild(self, parent, allfiles, changedfiles=None):
|
|
|
|
changedfiles = changedfiles or allfiles
|
|
|
|
oldmap = self._map
|
2007-08-06 06:04:56 +04:00
|
|
|
self.clear()
|
2013-03-07 08:13:09 +04:00
|
|
|
for f in allfiles:
|
|
|
|
if f not in changedfiles:
|
|
|
|
self._map[f] = oldmap[f]
|
2006-02-20 21:04:56 +03:00
|
|
|
else:
|
2013-03-07 08:13:09 +04:00
|
|
|
if 'x' in allfiles.flags(f):
|
2014-05-28 04:10:28 +04:00
|
|
|
self._map[f] = dirstatetuple('n', 0777, -1, 0)
|
2013-03-07 08:13:09 +04:00
|
|
|
else:
|
2014-05-28 04:10:28 +04:00
|
|
|
self._map[f] = dirstatetuple('n', 0666, -1, 0)
|
2007-06-18 22:24:34 +04:00
|
|
|
self._pl = (parent, nullid)
|
2007-07-22 01:02:09 +04:00
|
|
|
self._dirty = True
|
2005-08-28 01:21:25 +04:00
|
|
|
|
|
|
|
def write(self):
|
2007-06-18 22:24:34 +04:00
|
|
|
if not self._dirty:
|
2006-02-23 04:17:08 +03:00
|
|
|
return
|
2014-07-22 18:59:30 +04:00
|
|
|
|
|
|
|
# enough 'delaywrite' prevents 'pack_dirstate' from dropping
|
|
|
|
# timestamp of each entries in dirstate, because of 'now > mtime'
|
|
|
|
delaywrite = self._ui.configint('debug', 'dirstate.delaywrite', 0)
|
|
|
|
if delaywrite:
|
|
|
|
import time # to avoid useless import
|
|
|
|
time.sleep(delaywrite)
|
|
|
|
|
2008-03-19 23:55:21 +03:00
|
|
|
st = self._opener("dirstate", "w", atomictemp=True)
|
2009-10-01 19:17:52 +04:00
|
|
|
# use the modification time of the newly created temporary file as the
|
|
|
|
# filesystem's notion of 'now'
|
2012-05-30 23:55:33 +04:00
|
|
|
now = util.fstat(st).st_mtime
|
2013-04-28 01:19:52 +04:00
|
|
|
st.write(parsers.pack_dirstate(self._map, self._copymap, self._pl, now))
|
|
|
|
st.close()
|
|
|
|
self._lastnormaltime = 0
|
|
|
|
self._dirty = self._dirtypl = False
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2008-02-08 23:07:55 +03:00
|
|
|
def _dirignore(self, f):
|
2008-04-05 20:15:04 +04:00
|
|
|
if f == '.':
|
|
|
|
return False
|
2008-02-08 23:07:55 +03:00
|
|
|
if self._ignore(f):
|
|
|
|
return True
|
2013-04-11 02:08:25 +04:00
|
|
|
for p in scmutil.finddirs(f):
|
2008-07-12 03:46:02 +04:00
|
|
|
if self._ignore(p):
|
2008-02-08 23:07:55 +03:00
|
|
|
return True
|
|
|
|
return False
|
|
|
|
|
2013-05-07 21:02:55 +04:00
|
|
|
def _walkexplicit(self, match, subrepos):
|
|
|
|
'''Get stat data about the files explicitly specified by match.
|
2006-10-27 08:54:24 +04:00
|
|
|
|
2013-05-08 01:19:04 +04:00
|
|
|
Return a triple (results, dirsfound, dirsnotfound).
|
2013-05-07 21:02:55 +04:00
|
|
|
- results is a mapping from filename to stat result. It also contains
|
|
|
|
listings mapping subrepos and .hg to None.
|
2013-05-08 01:19:04 +04:00
|
|
|
- dirsfound is a list of files found to be directories.
|
2013-05-07 21:02:55 +04:00
|
|
|
- dirsnotfound is a list of files that the dirstate thinks are
|
|
|
|
directories and that were not found.'''
|
2008-05-12 20:37:07 +04:00
|
|
|
|
2009-06-01 02:54:18 +04:00
|
|
|
def badtype(mode):
|
2009-05-08 09:54:00 +04:00
|
|
|
kind = _('unknown')
|
2010-01-25 09:05:27 +03:00
|
|
|
if stat.S_ISCHR(mode):
|
|
|
|
kind = _('character device')
|
|
|
|
elif stat.S_ISBLK(mode):
|
|
|
|
kind = _('block device')
|
|
|
|
elif stat.S_ISFIFO(mode):
|
|
|
|
kind = _('fifo')
|
|
|
|
elif stat.S_ISSOCK(mode):
|
|
|
|
kind = _('socket')
|
|
|
|
elif stat.S_ISDIR(mode):
|
|
|
|
kind = _('directory')
|
2009-06-01 02:54:18 +04:00
|
|
|
return _('unsupported file type (type is %s)') % kind
|
2008-07-22 22:03:23 +04:00
|
|
|
|
2013-05-04 01:39:28 +04:00
|
|
|
matchedir = match.explicitdir
|
2009-06-01 02:54:18 +04:00
|
|
|
badfn = match.bad
|
2008-07-22 22:03:24 +04:00
|
|
|
dmap = self._map
|
2007-07-26 21:02:58 +04:00
|
|
|
normpath = util.normpath
|
|
|
|
lstat = os.lstat
|
2008-07-22 22:03:23 +04:00
|
|
|
getkind = stat.S_IFMT
|
2008-07-22 22:03:20 +04:00
|
|
|
dirkind = stat.S_IFDIR
|
2008-07-22 22:03:23 +04:00
|
|
|
regkind = stat.S_IFREG
|
|
|
|
lnkkind = stat.S_IFLNK
|
2008-07-22 22:03:24 +04:00
|
|
|
join = self._join
|
2013-05-08 01:19:04 +04:00
|
|
|
dirsfound = []
|
|
|
|
foundadd = dirsfound.append
|
2013-05-07 21:02:55 +04:00
|
|
|
dirsnotfound = []
|
2013-05-08 01:20:34 +04:00
|
|
|
notfoundadd = dirsnotfound.append
|
2008-07-22 22:03:10 +04:00
|
|
|
|
2013-05-07 21:02:55 +04:00
|
|
|
if match.matchfn != match.exact and self._checkcase:
|
2010-11-01 22:18:42 +03:00
|
|
|
normalize = self._normalize
|
|
|
|
else:
|
2012-12-04 22:29:18 +04:00
|
|
|
normalize = None
|
2010-11-01 22:18:42 +03:00
|
|
|
|
2010-09-11 01:53:51 +04:00
|
|
|
files = sorted(match.files())
|
|
|
|
subrepos.sort()
|
|
|
|
i, j = 0, 0
|
|
|
|
while i < len(files) and j < len(subrepos):
|
|
|
|
subpath = subrepos[j] + "/"
|
2011-01-04 14:53:11 +03:00
|
|
|
if files[i] < subpath:
|
2010-09-11 01:53:51 +04:00
|
|
|
i += 1
|
|
|
|
continue
|
2011-02-04 11:05:23 +03:00
|
|
|
while i < len(files) and files[i].startswith(subpath):
|
2010-09-11 01:53:51 +04:00
|
|
|
del files[i]
|
|
|
|
j += 1
|
|
|
|
|
2008-07-22 22:03:24 +04:00
|
|
|
if not files or '.' in files:
|
|
|
|
files = ['']
|
2010-01-01 02:19:30 +03:00
|
|
|
results = dict.fromkeys(subrepos)
|
|
|
|
results['.hg'] = None
|
2007-07-26 21:02:58 +04:00
|
|
|
|
2010-09-11 01:53:51 +04:00
|
|
|
for ff in files:
|
2012-12-04 22:29:18 +04:00
|
|
|
if normalize:
|
|
|
|
nf = normalize(normpath(ff), False, True)
|
|
|
|
else:
|
|
|
|
nf = normpath(ff)
|
2008-07-22 22:03:21 +04:00
|
|
|
if nf in results:
|
2008-07-22 22:03:10 +04:00
|
|
|
continue
|
|
|
|
|
|
|
|
try:
|
2008-07-22 22:03:24 +04:00
|
|
|
st = lstat(join(nf))
|
2008-07-22 22:03:23 +04:00
|
|
|
kind = getkind(st.st_mode)
|
|
|
|
if kind == dirkind:
|
2009-05-14 12:50:45 +04:00
|
|
|
if nf in dmap:
|
2013-10-03 20:01:21 +04:00
|
|
|
# file replaced by dir on disk but still in dirstate
|
2009-05-14 12:50:45 +04:00
|
|
|
results[nf] = None
|
2013-05-04 01:41:58 +04:00
|
|
|
if matchedir:
|
|
|
|
matchedir(nf)
|
2013-05-08 01:19:04 +04:00
|
|
|
foundadd(nf)
|
2010-09-24 21:46:54 +04:00
|
|
|
elif kind == regkind or kind == lnkkind:
|
2008-07-22 22:03:23 +04:00
|
|
|
results[nf] = st
|
2008-07-22 22:03:20 +04:00
|
|
|
else:
|
2009-06-01 02:54:18 +04:00
|
|
|
badfn(ff, badtype(kind))
|
2008-07-22 22:03:23 +04:00
|
|
|
if nf in dmap:
|
2008-07-22 22:03:21 +04:00
|
|
|
results[nf] = None
|
2013-10-03 20:01:21 +04:00
|
|
|
except OSError, inst: # nf not found on disk - it is dirstate only
|
|
|
|
if nf in dmap: # does it exactly match a missing file?
|
2009-06-01 02:54:18 +04:00
|
|
|
results[nf] = None
|
2013-10-03 20:01:21 +04:00
|
|
|
else: # does it match a missing directory?
|
2009-06-01 02:54:18 +04:00
|
|
|
prefix = nf + "/"
|
|
|
|
for fn in dmap:
|
|
|
|
if fn.startswith(prefix):
|
2013-05-04 01:41:58 +04:00
|
|
|
if matchedir:
|
|
|
|
matchedir(nf)
|
2013-05-08 01:20:34 +04:00
|
|
|
notfoundadd(nf)
|
2009-06-01 02:54:18 +04:00
|
|
|
break
|
2009-06-01 02:54:18 +04:00
|
|
|
else:
|
2009-06-01 02:54:18 +04:00
|
|
|
badfn(ff, inst.strerror)
|
2013-05-07 21:02:55 +04:00
|
|
|
|
2013-05-08 01:19:04 +04:00
|
|
|
return results, dirsfound, dirsnotfound
|
2013-05-07 21:02:55 +04:00
|
|
|
|
2013-04-23 04:11:18 +04:00
|
|
|
def walk(self, match, subrepos, unknown, ignored, full=True):
|
2013-05-07 21:02:55 +04:00
|
|
|
'''
|
|
|
|
Walk recursively through the directory tree, finding all files
|
|
|
|
matched by match.
|
|
|
|
|
2013-04-23 04:11:18 +04:00
|
|
|
If full is False, maybe skip some known-clean files.
|
|
|
|
|
2013-05-07 21:02:55 +04:00
|
|
|
Return a dict mapping filename to stat-like object (either
|
|
|
|
mercurial.osutil.stat instance or return value of os.stat()).
|
2013-04-23 04:11:18 +04:00
|
|
|
|
2013-05-07 21:02:55 +04:00
|
|
|
'''
|
2013-04-23 04:11:18 +04:00
|
|
|
# full is a flag that extensions that hook into walk can use -- this
|
|
|
|
# implementation doesn't use it at all. This satisfies the contract
|
|
|
|
# because we only guarantee a "maybe".
|
2013-05-07 21:02:55 +04:00
|
|
|
|
|
|
|
if ignored:
|
|
|
|
ignore = util.never
|
|
|
|
dirignore = util.never
|
2013-10-03 20:01:21 +04:00
|
|
|
elif unknown:
|
|
|
|
ignore = self._ignore
|
|
|
|
dirignore = self._dirignore
|
|
|
|
else:
|
|
|
|
# if not unknown and not ignored, drop dir recursion and step 2
|
2013-05-07 21:02:55 +04:00
|
|
|
ignore = util.always
|
|
|
|
dirignore = util.always
|
|
|
|
|
|
|
|
matchfn = match.matchfn
|
|
|
|
matchalways = match.always()
|
|
|
|
matchtdir = match.traversedir
|
|
|
|
dmap = self._map
|
|
|
|
listdir = osutil.listdir
|
|
|
|
lstat = os.lstat
|
|
|
|
dirkind = stat.S_IFDIR
|
|
|
|
regkind = stat.S_IFREG
|
|
|
|
lnkkind = stat.S_IFLNK
|
|
|
|
join = self._join
|
|
|
|
|
|
|
|
exact = skipstep3 = False
|
|
|
|
if matchfn == match.exact: # match.exact
|
|
|
|
exact = True
|
|
|
|
dirignore = util.always # skip step 2
|
|
|
|
elif match.files() and not match.anypats(): # match.match, no patterns
|
|
|
|
skipstep3 = True
|
|
|
|
|
|
|
|
if not exact and self._checkcase:
|
|
|
|
normalize = self._normalize
|
|
|
|
skipstep3 = False
|
|
|
|
else:
|
|
|
|
normalize = None
|
|
|
|
|
|
|
|
# step 1: find all explicit files
|
|
|
|
results, work, dirsnotfound = self._walkexplicit(match, subrepos)
|
2008-07-22 22:03:10 +04:00
|
|
|
|
2013-05-07 20:31:00 +04:00
|
|
|
skipstep3 = skipstep3 and not (work or dirsnotfound)
|
2013-05-07 20:47:10 +04:00
|
|
|
work = [d for d in work if not dirignore(d)]
|
|
|
|
wadd = work.append
|
|
|
|
|
2008-07-22 22:03:18 +04:00
|
|
|
# step 2: visit subdirectories
|
|
|
|
while work:
|
|
|
|
nd = work.pop()
|
2008-10-15 16:52:27 +04:00
|
|
|
skip = None
|
2008-07-22 22:03:18 +04:00
|
|
|
if nd == '.':
|
|
|
|
nd = ''
|
|
|
|
else:
|
2008-10-15 16:52:27 +04:00
|
|
|
skip = '.hg'
|
|
|
|
try:
|
|
|
|
entries = listdir(join(nd), stat=True, skip=skip)
|
|
|
|
except OSError, inst:
|
2012-10-24 08:25:22 +04:00
|
|
|
if inst.errno in (errno.EACCES, errno.ENOENT):
|
2013-10-03 20:01:21 +04:00
|
|
|
match.bad(self.pathto(nd), inst.strerror)
|
2008-10-15 16:52:27 +04:00
|
|
|
continue
|
|
|
|
raise
|
2008-07-22 22:03:18 +04:00
|
|
|
for f, kind, st in entries:
|
2012-12-04 22:29:18 +04:00
|
|
|
if normalize:
|
|
|
|
nf = normalize(nd and (nd + "/" + f) or f, True, True)
|
|
|
|
else:
|
|
|
|
nf = nd and (nd + "/" + f) or f
|
2008-07-22 22:03:21 +04:00
|
|
|
if nf not in results:
|
2008-07-22 22:03:20 +04:00
|
|
|
if kind == dirkind:
|
|
|
|
if not ignore(nf):
|
2013-05-04 01:41:58 +04:00
|
|
|
if matchtdir:
|
|
|
|
matchtdir(nf)
|
2008-07-22 22:03:20 +04:00
|
|
|
wadd(nf)
|
2013-03-23 04:03:00 +04:00
|
|
|
if nf in dmap and (matchalways or matchfn(nf)):
|
2008-07-22 22:03:21 +04:00
|
|
|
results[nf] = None
|
2008-07-22 22:03:25 +04:00
|
|
|
elif kind == regkind or kind == lnkkind:
|
|
|
|
if nf in dmap:
|
2013-03-23 04:03:00 +04:00
|
|
|
if matchalways or matchfn(nf):
|
2008-07-22 22:03:25 +04:00
|
|
|
results[nf] = st
|
2013-03-23 04:03:00 +04:00
|
|
|
elif (matchalways or matchfn(nf)) and not ignore(nf):
|
2008-07-22 22:03:21 +04:00
|
|
|
results[nf] = st
|
2013-03-23 04:03:00 +04:00
|
|
|
elif nf in dmap and (matchalways or matchfn(nf)):
|
2008-07-22 22:03:25 +04:00
|
|
|
results[nf] = None
|
2008-07-22 22:03:18 +04:00
|
|
|
|
2013-03-26 01:12:39 +04:00
|
|
|
for s in subrepos:
|
|
|
|
del results[s]
|
|
|
|
del results['.hg']
|
|
|
|
|
2013-10-03 20:01:21 +04:00
|
|
|
# step 3: visit remaining files from dmap
|
2009-06-01 02:54:18 +04:00
|
|
|
if not skipstep3 and not exact:
|
2013-10-03 20:01:21 +04:00
|
|
|
# If a dmap file is not in results yet, it was either
|
|
|
|
# a) not matching matchfn b) ignored, c) missing, or d) under a
|
|
|
|
# symlink directory.
|
2013-03-23 04:03:49 +04:00
|
|
|
if not results and matchalways:
|
|
|
|
visit = dmap.keys()
|
|
|
|
else:
|
|
|
|
visit = [f for f in dmap if f not in results and matchfn(f)]
|
|
|
|
visit.sort()
|
|
|
|
|
2013-02-05 02:27:15 +04:00
|
|
|
if unknown:
|
2013-10-03 20:01:21 +04:00
|
|
|
# unknown == True means we walked all dirs under the roots
|
|
|
|
# that wasn't ignored, and everything that matched was stat'ed
|
|
|
|
# and is already in results.
|
|
|
|
# The rest must thus be ignored or under a symlink.
|
2013-11-07 03:19:04 +04:00
|
|
|
audit_path = pathutil.pathauditor(self._root)
|
2013-02-05 02:27:15 +04:00
|
|
|
|
|
|
|
for nf in iter(visit):
|
|
|
|
# Report ignored items in the dmap as long as they are not
|
|
|
|
# under a symlink directory.
|
2013-05-03 20:44:50 +04:00
|
|
|
if audit_path.check(nf):
|
2013-02-11 00:23:39 +04:00
|
|
|
try:
|
|
|
|
results[nf] = lstat(join(nf))
|
2013-10-03 20:01:21 +04:00
|
|
|
# file was just ignored, no links, and exists
|
2013-02-11 00:23:39 +04:00
|
|
|
except OSError:
|
|
|
|
# file doesn't exist
|
|
|
|
results[nf] = None
|
2013-02-05 02:27:15 +04:00
|
|
|
else:
|
|
|
|
# It's either missing or under a symlink directory
|
2013-10-03 20:01:21 +04:00
|
|
|
# which we in this case report as missing
|
2013-02-05 02:27:15 +04:00
|
|
|
results[nf] = None
|
|
|
|
else:
|
|
|
|
# We may not have walked the full directory tree above,
|
2013-10-03 20:01:21 +04:00
|
|
|
# so stat and check everything we missed.
|
2013-02-05 02:27:15 +04:00
|
|
|
nf = iter(visit).next
|
|
|
|
for st in util.statfiles([join(i) for i in visit]):
|
|
|
|
results[nf()] = st
|
2008-07-22 22:03:21 +04:00
|
|
|
return results
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2010-01-01 02:19:30 +03:00
|
|
|
def status(self, match, subrepos, ignored, clean, unknown):
|
2009-10-01 23:36:45 +04:00
|
|
|
'''Determine the status of the working copy relative to the
|
|
|
|
dirstate and return a tuple of lists (unsure, modified, added,
|
|
|
|
removed, deleted, unknown, ignored, clean), where:
|
|
|
|
|
|
|
|
unsure:
|
|
|
|
files that might have been modified since the dirstate was
|
|
|
|
written, but need to be read to be sure (size is the same
|
|
|
|
but mtime differs)
|
|
|
|
modified:
|
|
|
|
files that have definitely been modified since the dirstate
|
|
|
|
was written (different size or mode)
|
|
|
|
added:
|
|
|
|
files that have been explicitly added with hg add
|
|
|
|
removed:
|
|
|
|
files that have been explicitly removed with hg remove
|
|
|
|
deleted:
|
|
|
|
files that have been deleted through other means ("missing")
|
|
|
|
unknown:
|
|
|
|
files not in the dirstate that are not ignored
|
|
|
|
ignored:
|
|
|
|
files not in the dirstate that are ignored
|
|
|
|
(by _dirignore())
|
|
|
|
clean:
|
|
|
|
files that have definitely not been modified since the
|
|
|
|
dirstate was written
|
|
|
|
'''
|
2008-06-26 23:35:50 +04:00
|
|
|
listignored, listclean, listunknown = ignored, clean, unknown
|
2006-03-30 00:58:34 +04:00
|
|
|
lookup, modified, added, unknown, ignored = [], [], [], [], []
|
2006-07-21 03:21:07 +04:00
|
|
|
removed, deleted, clean = [], [], []
|
2005-08-28 01:21:25 +04:00
|
|
|
|
2007-07-26 21:02:58 +04:00
|
|
|
dmap = self._map
|
2009-10-01 23:36:45 +04:00
|
|
|
ladd = lookup.append # aka "unsure"
|
2007-07-26 21:02:58 +04:00
|
|
|
madd = modified.append
|
|
|
|
aadd = added.append
|
|
|
|
uadd = unknown.append
|
|
|
|
iadd = ignored.append
|
|
|
|
radd = removed.append
|
|
|
|
dadd = deleted.append
|
|
|
|
cadd = clean.append
|
2012-12-04 02:21:45 +04:00
|
|
|
mexact = match.exact
|
|
|
|
dirignore = self._dirignore
|
|
|
|
checkexec = self._checkexec
|
|
|
|
copymap = self._copymap
|
|
|
|
lastnormaltime = self._lastnormaltime
|
2007-07-26 21:02:58 +04:00
|
|
|
|
2013-04-24 01:16:33 +04:00
|
|
|
# We need to do full walks when either
|
|
|
|
# - we're listing all clean files, or
|
|
|
|
# - match.traversedir does something, because match.traversedir should
|
|
|
|
# be called for every dir in the working dir
|
|
|
|
full = listclean or match.traversedir is not None
|
|
|
|
for fn, st in self.walk(match, subrepos, listunknown, listignored,
|
|
|
|
full=full).iteritems():
|
2008-05-12 20:37:08 +04:00
|
|
|
if fn not in dmap:
|
2012-12-04 02:21:45 +04:00
|
|
|
if (listignored or mexact(fn)) and dirignore(fn):
|
2008-06-26 23:35:50 +04:00
|
|
|
if listignored:
|
2008-02-08 23:07:55 +03:00
|
|
|
iadd(fn)
|
2013-10-14 08:25:29 +04:00
|
|
|
else:
|
2007-07-26 21:02:58 +04:00
|
|
|
uadd(fn)
|
2005-10-28 00:29:35 +04:00
|
|
|
continue
|
2008-05-12 20:37:08 +04:00
|
|
|
|
2014-05-28 08:02:16 +04:00
|
|
|
# This is equivalent to 'state, mode, size, time = dmap[fn]' but not
|
|
|
|
# written like that for performance reasons. dmap[fn] is not a
|
|
|
|
# Python tuple in compiled builds. The CPython UNPACK_SEQUENCE
|
|
|
|
# opcode has fast paths when the value to be unpacked is a tuple or
|
|
|
|
# a list, but falls back to creating a full-fledged iterator in
|
|
|
|
# general. That is much slower than simply accessing and storing the
|
|
|
|
# tuple members one by one.
|
|
|
|
t = dmap[fn]
|
|
|
|
state = t[0]
|
|
|
|
mode = t[1]
|
|
|
|
size = t[2]
|
|
|
|
time = t[3]
|
2008-05-12 20:37:08 +04:00
|
|
|
|
2008-07-22 22:02:36 +04:00
|
|
|
if not st and state in "nma":
|
|
|
|
dadd(fn)
|
|
|
|
elif state == 'n':
|
2011-03-25 17:03:53 +03:00
|
|
|
mtime = int(st.st_mtime)
|
2008-03-14 15:56:58 +03:00
|
|
|
if (size >= 0 and
|
2012-10-09 02:50:42 +04:00
|
|
|
((size != st.st_size and size != st.st_size & _rangemask)
|
2012-12-04 02:21:45 +04:00
|
|
|
or ((mode ^ st.st_mode) & 0100 and checkexec))
|
2010-04-20 13:17:01 +04:00
|
|
|
or size == -2 # other parent
|
2012-12-04 02:21:45 +04:00
|
|
|
or fn in copymap):
|
2007-07-26 21:02:58 +04:00
|
|
|
madd(fn)
|
2013-08-31 21:20:15 +04:00
|
|
|
elif time != mtime and time != mtime & _rangemask:
|
2007-07-26 21:02:58 +04:00
|
|
|
ladd(fn)
|
2012-12-04 02:21:45 +04:00
|
|
|
elif mtime == lastnormaltime:
|
2011-03-25 17:03:53 +03:00
|
|
|
# fn may have been changed in the same timeslot without
|
|
|
|
# changing its size. This can happen if we quickly do
|
|
|
|
# multiple commits in a single transaction.
|
|
|
|
# Force lookup, so we don't miss such a racy file change.
|
dirstate: avoid a race with multiple commits in the same process
(issue2264, issue2516)
The race happens when two commits in a row change the same file
without changing its size, *if* those two commits happen in the same
second in the same process while holding the same repo lock. For
example:
commit 1:
M a
M b
commit 2: # same process, same second, same repo lock
M b # modify b without changing its size
M c
This first manifested in transplant, which is the most common way to
do multiple commits in the same process. But it can manifest in any
script or extension that does multiple commits under the same repo
lock. (Thus, the test script tests both transplant and a custom script.)
The problem was that dirstate.status() failed to notice the change to
b when localrepo is about to do the second commit, meaning that change
gets left in the working directory. In the context of transplant, that
means either a crash ("RuntimeError: nothing committed after
transplant") or a silently inaccurate transplant, depending on whether
any other files were modified by the second transplanted changeset.
The fix is to make status() work a little harder when we have
previously marked files as clean (state 'normal') in the same process.
Specifically, dirstate.normal() adds files to self._lastnormal, and
other state-changing methods remove them. Then dirstate.status() puts
any files in self._lastnormal into state 'lookup', which will make
localrepository.status() read file contents to see if it has really
changed. So we pay a small performance penalty for the second (and
subsequent) commits in the same process, without affecting the common
case. Anything that does lots of status updates and checks in the
same process could suffer a performance hit.
Incidentally, there is a simpler fix: call dirstate.normallookup() on
every file updated by commit() at the end of the commit. The trouble
with that solution is that it imposes a performance penalty on the
common case: it means the next status-dependent hg command after every
"hg commit" will be a little bit slower. The patch here is more
complex, but only affects performance for the uncommon case.
2011-03-21 00:41:09 +03:00
|
|
|
ladd(fn)
|
2008-06-26 23:35:50 +04:00
|
|
|
elif listclean:
|
2007-07-26 21:02:58 +04:00
|
|
|
cadd(fn)
|
2008-05-12 20:37:08 +04:00
|
|
|
elif state == 'm':
|
2007-07-26 21:02:58 +04:00
|
|
|
madd(fn)
|
2008-05-12 20:37:08 +04:00
|
|
|
elif state == 'a':
|
2007-07-26 21:02:58 +04:00
|
|
|
aadd(fn)
|
2008-05-12 20:37:08 +04:00
|
|
|
elif state == 'r':
|
2007-07-26 21:02:58 +04:00
|
|
|
radd(fn)
|
2005-10-28 00:29:35 +04:00
|
|
|
|
2006-07-21 03:21:07 +04:00
|
|
|
return (lookup, modified, added, removed, deleted, unknown, ignored,
|
|
|
|
clean)
|
2014-08-02 09:05:16 +04:00
|
|
|
|
|
|
|
def matches(self, match):
|
|
|
|
'''
|
|
|
|
return files in the dirstate (in whatever state) filtered by match
|
|
|
|
'''
|
|
|
|
dmap = self._map
|
|
|
|
if match.always():
|
|
|
|
return dmap.keys()
|
|
|
|
files = match.files()
|
|
|
|
if match.matchfn == match.exact:
|
|
|
|
# fast path -- filter the other way around, since typically files is
|
|
|
|
# much smaller than dmap
|
|
|
|
return [f for f in files if f in dmap]
|
|
|
|
if not match.anypats() and util.all(fn in dmap for fn in files):
|
|
|
|
# fast path -- all the values are known to be files, so just return
|
|
|
|
# that
|
|
|
|
return list(files)
|
|
|
|
return [f for f in dmap if match(f)]
|